The basic idea is that if we don't find a variable `foo` then we fallback to retrieving the value from the tiddler `$:/global/foo`, if it exists.
This allows us to replace the usual importvariables-based mechanism for global definitions, avoiding cluttering up the variable namespace with every macro.
In order to permit subprocedures to be overridden, we also introduce a mechanism for conditional definitions: preceding the word definition|procedure|function|widget with a + causes the definition only to occur if the specified variable doesn't already exist. In the next commit we'll apply this mechanism to the tabs macro
* Extend HTML tag parser to maintain an ordered array of attribute names
* Add some tests for repeated attributes
* Record entire attribute in orderedAttributes array so that we can work with duplicated attributes
* html-comment, that can be used in the pragma area
* add commentpragma test
* fix typo
* fix typo and change comments ab bit
* combine html-comment and pragma-comment and add some docs, how to use it
* Make docs simpler by removing caching info
* change h2 wording
We can now pass arrays of rule classes to the parser constructor, overriding the rules that would normally be used by the parser.
This allows us to create custom variants of the wikitext parser with their own content type.
It could also provide a basis for a new Markdown parser based on our existing wikitext parser but with new rules.
* wikirules now use better macrocall parser
Before, wikirules would use a deficient macrocall parser which couldn't
handle certain types of arguments. Now it uses the same one that the
widget parser uses. Less code!
* style changes and removing weird switch statement
That switch statement made more sense in an earlier iteration.
* comment improvements
* oops, wikirule macrocalls could do ONE thing better
* '=' wasn't allowed for widget macros, but why?
Now they're allowed for both widget macros and macrocall macros.
This fixes a problem introduced in
c7b31b0242.
The changes by @tobibeer inadvertently made the regular expression
evaluation significantly more expensive because of lookahead. The is
less elegant but reverts the performance problem.
